compute-desk

Availability Sets — FDs, UDs, when AS still applies

VM AZ design — zonal vs zone-redundant, latency, AZ-aware LB, SLA

Capacity Reservation Groups (CRG) — guarantee capacity for critical/DR VMs

Azure Dedicated Hosts — host groups, isolation, BYOL, maintenance control

Proximity Placement Groups — co-locate for ultra-low latency

Azure Scheduled Events via IMDS — react to host maintenance

Compose Azure VM SLAs — single-VM, AS, AZ, multi-region

Configure Azure Backup for VMs

Cross-Region Restore (CRR)

Vault immutability and soft delete — ransomware posture

Restore VM, disk, or file — OLR/ALR/instant/cross-zone/cross-region

Daily/weekly/monthly/yearly retention — GFS, archive tier

Standard vs Enhanced Recovery Services Vault

Break down VM cost in Cost Management — meters, grouping, KQL queries

Azure Dev/Test pricing — eligibility and benefits

Compare PAYG vs Reserved Instance vs Savings Plan vs Spot

Azure Reserved VM Instances — 1y vs 3y, scope, instance size flexibility, exchange/refund rules

Azure Compute Savings Plan — hourly commitment, family/region flexibility

Spot VMs and Spot VMSS — eviction policy, max price, workload fit

ASR architecture for Azure-to-Azure VM replication

ASR network mapping — IP retention, DNS cutover, NSG/UDR re-attachment

Cross-region replication options — ASR vs ZRS vs cross-region zonal restore

Test failover without disruption — isolated VNet, checklist, cadence

ASR recovery plans — multi-VM ordering, scripts, manual gates

Translate business RTO/RPO into ASR config + architecture

Understand on-demand vs credit-based bursting on Premium SSDs

Choose between PMK, CMK, double encryption, and encryption at host

Compare Standard HDD, Standard SSD, Premium SSD v1/v2, and Ultra Disk

Use local-SSD-backed ephemeral OS disks for stateless workloads

Set ReadOnly, ReadWrite, or None host caching per disk role

Use Premium SSD v2 for cost-efficient high IOPS

Attach a single managed disk to multiple VMs (clustered apps)

Configure Ultra Disks for extreme IOPS / low latency workloads

Pick the right N-series SKU (NC, ND, NV) for training / inference / visualization

Design H-series HPC clusters (HBv3/HBv4, HX) with InfiniBand

Configure SR-IOV InfiniBand and verify RDMA bandwidth

Use Azure ML / DSVM / NGC images for ready-to-run training stacks

Stand up MPI (OpenMPI, Intel MPI, HPC-X) clusters on VMSS

Install NVIDIA drivers + CUDA via extension or HPC image

Deploy Azure Bastion (Basic, Standard, Premium) for browser-based SSH/RDP